CAC instability = system instability
The brands that have earned the ability to scale aren’t the best media buyers, they’re the ones feeding the best signals into their ecommerce growth machine.
It’s obvious to say that most founders want to scale, however for many the cost of acquiring new customers becomes prohibitive. So you fix the ads and diagnose that a Meta problem… or a Google problem, but nothing resolves the CAC problem.
And I keep saying this all day long (to clients, not just myself…):
“scaling isn’t an action… it’s an outcome”. Let me explain.
Scaling exposes weaknesses. It doesn’t fix them.
When it comes to scaling with confidence, they’re probably looking at the wrong thing. The channel itself is rarely the problem. The problem is what’s being fed into it. The inputs are weak, the signals are noisy, and the machine is working exactly as designed. It’s just designed on a foundation of poor information.
That’s an input problem. And an input problem is a scaling problem.
The ads channel itself, whether that be Google or Meta isn’t broken.
The channel is faithfully executing on your bad inputs.
I’ve worked with brands spending serious money on paid media who couldn’t tell me why their customer bought. Couldn’t explain what actually made someone choose them. Hadn’t tested their messaging in any meaningful way. And yet they were convinced the issue was their bidding strategy.
It wasn’t. Fix the inputs. The outputs follow. Stick with me on this one. What I’m sharing is key to your ecommerce growth success. Let’s start out by taking a look at what’s now changed.
Old World vs New World
Not long ago, media buying was a craft, and at the heart of value that an Ads Agency delivered to you. They built complex campaign structures, sat on top of data so they could set bids manually, and carved audiences by hand. The master skill was in the configuration. The more granular you went, the more control you had. Or so they told themselves and their clients.
That world is mostly gone.
The platforms have absorbed those levers;
- Meta’s algorithm decides who sees your ad.
- Google’s Performance Max decides where it runs.
The targeting, the bidding, the placement, all of it is now processed by machine learning systems that are faster, more adaptive and handling more variables than any human operator. They’re damn smart.
What this means to you as you scale your brand is fundamental.
In the old world, human control of the platform structure was the lever. In the new world, the lever is what you feed the platform. The AI does the processing. You provide the inputs.
You’re no longer simply managing the machine.
You’re feeding it. The question is whether what you’re feeding it is any good.
This is the shift that most brands haven’t (yet) made. They’re still optimising for control of a system that no longer wants to be controlled. They’re adjusting budgets, tweaking audiences, reshuffling campaigns, when the actual work is happening upstream, before any of that.
So in Ads Speak, What Exactly Is The Input Layer?
The Input Layer is everything that feeds your growth system.
It sits before the platforms. Before the campaigns. Before the creative goes live. It’s the quality of understanding, information and signal that you bring to your growth machine (in this case, Meta or Google’s growth machine).
Think of it in three core categories, or more specifically, signals:
1.) You deliver Customer signals.
What your customers actually think, feel, say and do. Their language. Their hesitations. Their reasons for buying and their reasons for leaving.
2.) You deliver Message signals.
How clearly and compellingly you articulate your value. The positioning. The creative. The offer. The copy that either resonates or doesn’t.
3.) You deliver Data signals.
The accuracy of your tracking, attribution and first-party data. What you’re actually telling the platforms about what’s working and who’s converting.
All three feed into the same machine. And all three can undermine it.
Let’s Take a Look Inside the Input Layer
1.) Customer Understanding
This is where almost every brand I work with has the biggest gap.
They know their demographics (eg. 35-44F). They don’t know their buyer.
- What’s the specific problem your customer has that your product solves?
- What language do they use when they describe it?
- What did they try before?
- What nearly stopped them from buying?
Without that, you’re writing copy (or AI is manufacturing your copy) in the dark. Your creative is a stab in the dark. Your messaging is generic. And the platform is learning from engagement signals that don’t reflect genuine intent.
2.) Messaging and Positioning
Most ecommerce messaging is product-led. Features, specifications, what’s in the box. That’s not what converts.
Conversion happens when a customer sees themselves in the message. When the copy reflects their situation, their frustration, their aspiration. When the brand sounds like it knows them.
Weak positioning is a weak input. The platform can amplify your reach but it cannot fix your message.
3.) Creative Signal Quality
Creative is now one of the primary signals you send to platforms. The algorithm learns from what people engage with, watch, click, and convert from.
If your creative is generic, the signal is generic. The platform learns to find people who respond to generic creative. That’s rarely your best customer, and one of the main reasons why conversion rate kills your CAC.
Strong creative, grounded in real customer language and insight, teaches the machine who to find. It’s signal, not just content.
4.) Tracking and Attribution
If your data is broken, you’re feeding the machine lies.
Misconfigured events. Missing purchase signals. Inflated click-through attribution. These are all inputs into a system that acts on what it receives. Feed it inaccurate conversion data and it optimises toward the wrong thing.
Tracking is not a technical afterthought. It’s a foundational input. This is where I find myself helping more and more brands – simply ensuring that server-side tracking is installed and feeding into platforms correct conversion data. It’s a simple fix to many scaling issues.
5.) First-Party Data
Your customer list is a signal. Your email subscribers are a signal. Your purchasers, your repeat buyers, your highest LTV cohorts, all signal.
When you feed these into the platforms as seed audiences, lookalikes and exclusions, you’re giving the machine a map of who you want. That map is only as good as the data behind it.
A strong first-party data set, properly structured and regularly updated, is one of the most valuable inputs you have.
6.) Offer Strength
Offer is an input. Not just a promotional mechanic.
A weak product description page (or product context page) creates friction at every stage. The click is reluctant. The landing page conversion is low. The algorithm sees poor performance and pulls back.
A strong offer, one that genuinely matches customer motivation, pulls through the entire system. It improves ad performance, conversion rate and post-purchase satisfaction simultaneously.
7.) Retention and LTV Signals
What happens after the first purchase is an input into what happens before the next one.
Your post-purchase sequence, your product experience, your email flows, all generate behavioural data that should be feeding back into your acquisition strategy. Which cohorts retain? Which channels produce buyers who come back? Platforms like RetentionX allow you to easily create valuable cohorts that can demonstrate to the ad platforms the exact customer profile you’re looking to attract.
Most brands don’t close this loop. They treat retention as separate from acquisition. They’re the same system.
8.) Feedback Loops
The machine improves when it learns. Learning requires feedback. And feedback requires you to capture it systematically.
Post-purchase surveys. Customer interviews. On-site behaviour data. Returns analysis. These are all inputs. They’re also the inputs most brands skip because they take effort.
That effort is exactly the work that separates a machine that improves over time from one that plateaus.
What Founders Actually Feel
You’d rather:
- Improve conversion by 20%
- Increase AOV
- Tighten targeting quality
…than just push more spend, right?
Because you know:
Every improvement compounds across the system
Whereas:
More spend compounds inefficiency
So, let’s tackle these economic challenges head on;
Rising CAC. That’s the most common complaint I hear. Costs are going up and results are getting harder to predict.
Poor conversion. Traffic is there but it’s not converting at the rate it used to.
Inconsistent performance. Good weeks and bad weeks with no clear explanation. The algorithm seems to be doing whatever it wants.
Ads not working. A catch-all for a system that isn’t producing the results the business needs.
Every one of these is an output.
They feel like channel problems. They’re described as channel problems. But they’re symptoms of a broken input layer.
Rising CAC is not a Meta problem. It’s a signal that the machine has less to work with than it needs.
When customer understanding is shallow, messaging is generic and tracking is unreliable, the platform is operating on poor inputs. It’s finding the wrong people, with the wrong message, and the data it’s learning from is distorted.
Of course performance is inconsistent.
The fix isn’t in the campaign. It’s upstream.
The Compounding Effect
Today’s AI-driven platforms scale upon what they’re given. They don’t have the mindset of an FD. They’re they’re to please by providing you what you ask for. And you’re NOT asking for just traffic.
- Strong inputs improve machine learning.
- Better signals produce better targeting.
- Better targeting produces better conversion data.
- Better conversion data improves the next round of targeting.
The loop compounds in your favour and weak inputs do the same thing. In the opposite direction.
- The machine learns from noise.
- Optimises toward poor-fit customers.
- Conversion suffers.
- The algorithm sees underperformance and becomes more cautious.
- Spend efficiency drops.
- CAC rises.
And as much as we’d like to point the finger and believe it is… it’s not a platform failure. That’s the machine doing exactly what it was designed to do with the information it was given.
AI doesn’t fix bad inputs. It scales them. Give it noise and it will find more noise, efficiently.
This is why the obsession with platform tactics is so costly. Every hour spent adjusting campaign structures, testing new objective settings, reshuffling budgets, is an hour not spent improving the inputs that actually determine what the machine produces.
Manage the Machine
The Manage the Machine framework is built on a simple idea. Growth is a system of connected parts, and your job as the operator is to design, feed and improve that system.
The machine has three layers.
Inputs. Customer understanding, messaging quality, creative signal, tracking accuracy, first-party data, offer strength, retention insight.
Processing. The platforms, the algorithms, the automation. This layer is increasingly outside your direct control. It runs on what you feed it.
Outputs. CAC, conversion rate, ROAS, LTV, revenue growth. The numbers your business lives and dies by.
The mistake most operators make is trying to manage the outputs by adjusting the processing layer. They pull campaign levers when the real problem is in the input layer.
Managing the machine isn’t about controlling platforms. It’s about controlling what feeds them. That’s your job as the machine operator.
Get the inputs right and the processing layer has what it needs to produce better outputs. That’s how you create predictable, scalable growth. Not by finding the right bidding strategy. By building a better machine.
Where to Start
The good news is that none of this requires a platform overhaul.
Instead, it requires a shift in where you focus your attention.
Stop jumping between tactics. Seriously. Tactic switching is what brands do when they don’t understand why things aren’t working. It creates the illusion of progress while the underlying problem remains untouched. Founders. In-house marketers. Agencies. We’re all culpable.
Audit your inputs. Work through the input layer systematically. Where is the understanding shallow? Where is the messaging generic? Where is the data broken? Most brands find the gaps quickly because they’ve never looked before.
Fix your tracking first. If your data signals are corrupted, everything else is guesswork. Clean tracking isn’t glamorous work. It’s the most important thing in the system.
Do the customer work. This is the input that compounds most over time. Real customer understanding improves every part of the machine simultaneously. Better ads. Better landing pages. Better email. Better retention. All from the same source.
Improve signal quality progressively. You don’t need to fix everything at once. Each improvement to the input layer produces a measurable improvement in outputs. Do the work in order of leverage.
| 5 Questions for you to now consider: 1.) If I asked my team why our best customer buys, would the answer come from data or assumption? 2.) Is my tracking accurate enough that I trust the conversion signals being sent to our ad platforms? 3.) When we’ve seen CAC rise, have we looked upstream, or have we adjusted bids? 4.) Do we have a systematic way to capture customer insight and feed it back into our messaging? 5.) Are we treating retention data as an input into our acquisition strategy? |
So… Who Are The Brands That Win in Ecommerce?
The brands winning in an AI-driven environment aren’t winning because they have the best media buyers. They’re winning because they’ve built better input layers.
They understand their customers deeply. They’ve tested their messaging rigorously. Their tracking is clean. Their creative is grounded in real insight. Their first-party data is structured and actively used.
They’re feeding the machine better signals. And the machine is rewarding them for it.
You don’t have a channel performance problem. You have an input problem. Fix what feeds the machine and the machine will do its job.
The work isn’t glamorous. It’s not the quick win that a new campaign structure promises. But it’s the work that compounds. The work that builds a growth system that keeps producing as you scale.
That’s the machine worth managing.
Written By:
